Security Consolidation in the Aftermath of Civil War: Explaining the Fates of Victorious Militias

Abstract

Policymakers and peacebuilding research often focus on rebel groups when studying demobilization and integration processes, but post-war governments must also manage the non-state militias that helped them gain or maintain power. Why do some post-war governments disintegrate their militia allies, while others integrate them into the military? We argue that when a salient ethnic difference exists between the (new) ruling elite and an allied militia, a process of mutual uncertainty in the post-war period will incentivize governments to disintegrate the group. However, governments will be most likely to integrate their militias when the military has sufficient coercive capabilities but few organizational hindrances to re-organizing. Using new data on the post-war fates of victorious militias across all civil conflicts from 1989 to 2014, we find robust support for these claims. The results suggest that a government’s optimal militia management strategy is shaped by both social and organizational constraints during the post-war period.

Keywords

militias civil wars domestic security military integration DDR

Introduction

The use of auxiliary non-state militias by governments is a controversial attribute of most contemporary civil wars. Such groups are often harbingers of social instability and frequently indulge in gratuitous violence, perpetrating some of the most egregious violations of human rights in contemporary history. Militias like the notorious Interahamwe in Rwanda and Janjaweed in Sudan are pervasive across time and space, yet scholars have recently remarked that “we have little systematic understanding of when and why the link between the militia and the state breaks off and what then happens to these groups…The death…or political ‘afterlife’ of militias is unexplored territory” (Carey and Mitchell 2017, 135). In the Democratic Republic of the Congo, for instance, the copious militia groups have met wildly different fates after aiding the government in combating rebel organizations: the government has integrated large portions of the Coalition of Congolese Patriotic Resistance but allowed many other Mai Mai village militias to remain active even in post-conflict periods. Elsewhere, governments intentionally form or enlist militias that are ethnically or religiously different from the ruling elites and then quickly dissemble them after the conflict. Examples of this pattern abound, from the Civil Defense Patrols of Guatemala, to the Comité de Vigilance de Tassara in Niger, to the various Tamil militias in Sri Lanka.

Despite the wealth of literature on militarily integrating (Glassmyer and Sambanis 2008; Krebs and Licklider 2015) or otherwise sharing power with opposition forces after a civil war (Hartzell and Hoddie 2003; Matanock 2016), current research often ignores the fact that rebels are not the only non-state armed actors that governments must deal with once violence has subsided (Staniland 2015). We take an important step toward filling this gap by investigating the fates of victorious militias in the aftermath of civil war. We define a victorious militia as an armed, non-state organization that fought as part of the victorious side in the war and/or explicitly coordinated with or was subservient to the ruling elite in charge of the executive after the conflict has ended.

Since governments tend to remain in power after civil wars more often than rebels seize power, many of these groups are pro-government militias (PGMs) with some form of linkage with the incumbent ruling elite. However, when rebel leaders take control of government or successfully secede, they too must manage their main non-state organization as well as any auxiliary forces that helped bring them to power. Regardless of whether the incumbents retain power or rebels take it, in the absence of the conflict in which militia forces were meant to fight, the new ruling elite must decide whether to dissolve or integrate these non-state forces that fought on their behalf or allow them to continue operating as unofficial pro-state militias.

We generated Figure 1 using our global sample (described below) of victorious non-state militia groups. Figure 1 illustrates two important points. First, victorious militias that were active in a civil war frequently survive beyond the cessation of violence. Second, our data exhibit remarkable variation in how these groups terminate. Particularly within the first five post-conflict years, victorious militias are disintegrated in some contexts and integrated into the military in many others. What explains this variation in the fates of victorious militias? Why are some militias who fought for the “victorious” side in a war integrated into the official security apparatus while others are dissolved?

Figure 1.

Militia survival after civil war.

We argue that post-war governments face difficult state-building and distributional challenges that may require terminating the militias that had helped bring them to victory in a civil war. Governments must then consider whether and how to manage these groups based on existing political, social, and organizational incentives. When a victorious militia has few social ties to the (new) ruling elite, a process of uncertainty and mutual commitment problems that characterize post-conflict situations, combined with little to no representation within the ruling cadre, will lead to a higher probability that the government eliminates its former allies altogether. Alternatively, when the post-war military has sufficient organizational resources to integrate these militia forces but also little organizational resistance to change and high enough uncertainty that it could exterminate the group in the future if necessary, the probability that the government formally integrates the militia into the official military apparatus will be greatest. This leads us to expect that the victorious militias are most likely to be absorbed into the military at intermediate levels of military organizational capacity. Using new data on victorious militia fates across all civil conflicts from 1989 to 2014, we find strong support for our hypotheses.

To our knowledge, this is one of the first systematic investigations of how, when, and why victorious non-state militias terminate after civil conflicts end. Beyond providing a nuanced theoretical argument to explain victorious militia fates as well as new, time-varying data on when militia integration and disintegration occurred, our analysis broadly suggests that states implement widely distinct strategies to manage these militias once civil war violence has subsided (Staniland 2015). This is important because failing to contain, appease, or effectively dissolve non-state groups after a civil war can pose problems for state-building (Reno 2011), incite recurrent conflict (Steinert, Steinert, and Carey 2019), and provide incentives for governments to engage in repressive practices or other human rights violations even during peacetime (Mitchell, Carey, and Butler 2014).

Militias and Auxiliary Forces

States and political elites frequently use different types of non-state auxiliary armed forces (e.g. PGMs, loyalist paramilitary units, civilian defense forces) to compete for power (Raleigh 2016), repress dissent (Carey and González 2020), avoid accountability for human rights atrocities (Carey, Colaresi, and Mitchell 2015), shield against coup threats (Carey, Colaresi, and Mitchell 2016), and supplement the military in counterinsurgent operations (Jentzsch, Kalyvas, and Schubiger 2015; Carey and Mitchell 2016; Peic 2014). A significant body of work classifies militias in terms of purpose (Raleigh 2016; Böhmelt and Clayton 2018), relationships with the state (Carey, Mitchell, and Lowe 2013; Staniland 2015), structure (Böhmelt and Clayton 2018), and composition (Abbs, Clayton, and Thomson 2020), and most do so within a principal-agent framework where the principal (state, elites, etc.) delegates some power or responsibilities to militia “agents” to obtain certain benefits. Many strategic benefits of militias—including coup-proofing, state-building, and repression—are applicable outside of civil war contexts, and in some cases, militias formed during a war evolve after a war to function as a more formal part of the security apparatus (Ahram 2011).

During periods of conflict, non-state armed forces, including pro-government and rebel-supporting militias, can provide various tactical benefits against the opposing side (Peic 2014; Jentzsch, Kalyvas, and Schubiger 2015; Carey and Mitchell 2016). Although some militias are better understood as private armies that help political leaders compete for power, many others are distinctly “emergency militias” that are formed explicitly as a supplement to regular security forces during civil conflict (Raleigh 2016). The organizational structures and capacity of these militias are largely influenced by their security responsibilities or purpose, their composition, and their links to the regime. Some serve as relatively cheap sources of additional firepower in conventional wars against insurgents, often leading to more organized, well-equipped militias with acknowledged, semiofficial ties to the regime (Jentzsch, Kalyvas, and Schubiger 2015; Carey and Mitchell 2016, 2017). When attached to certain autocrats, some of these militias may have a more favorable status in the eyes of the regime than the official military, which has important implications for structuring the post-war security apparatus (Ahram 2011; Carey, Colaresi, and Mitchell 2016).

In other cases, the regime will employ local civil defense militias to perform defensive tasks, hold territory, and obtain local intelligence that reduces rebel identification problems (Clayton and Thomson 2016; Kalyvas 2006). These groups are recruited from civilian populations, organizationally disaggregated, and often do not operate under traditional military-style command structures (Clayton and Thomson 2016). Unless they are meant to promote a national identity, they also tend to have informal (unacknowledged) links to the regime, which are associated with “recruitment procedures [that] are…lax, training more rudimentary, [and] discipline less predictable” (Carey and Mitchell 2017, 133). However, civilian militias can control civilians effectively in contested territory (Kalyvas 2006) and reveal private information about opposition groups to the principal, particularly when they are “composed from the anti-government population” or were “defector” groups from the rebellion itself (Abbs, Clayton, and Thomson 2020, 904; Kalyvas 2008; Lyall 2010).

Many rebel movements reap similar tactical benefits from auxiliary forces of their own. These subsidiary groups can be independent local militia that “operate to secure local control bases while allying with rebel groups seeking national power” (Raleigh 2016, 291), co-opted civil militia forces that switched sides (e.g. some People’s Militia units in Ethiopia), or auxiliaries explicitly created by the rebels to gather local intelligence or hold territory (e.g. the Mujibas in Mozambique or the Munshawa Hpyen Hpung in Burma). Thus, although most large-N analyses of militia politics focus on PGMs, different types of non-state militia forces can be active on either side of a war (Raleigh 2016).

Regardless, the literature on principal-agent models suggests that principals fear moral hazard problems, shirking, and agency slack when employing non-state armed forces. Militia organizations may perform poorly if members are not adequately compensated and monitored (Carey and Mitchell 2017), or they may switch sides entirely (Otto 2018). These problems are exacerbated when the threat from the opposition has subsided. Ethnic militias that assisted the government during a conflict, for instance, may have improved the government’s ability to defeat insurgents, but with no mutual threat to reinforce principal-agent cooperation, empowering these armed groups and failing to subsequently transform, appease, or eliminate them after the conflict could lead to domestic instability, renewed conflict (Steinert, Steinert, and Carey 2019), or state failure (Reno 2011). After a conflict ends, some counterinsurgent militias can evolve and serve a different purpose (Ahram 2011), but this is typically a strategy of the state to co-opt these forces to pursue its own long-term ideological objectives (Staniland 2015).

Despite the burgeoning literature on militias in civil wars and the challenges they pose to post-war governance, very little research has emphasized how governments deal with their non-state armed partners after conflicts end. This is surprising considering the wealth of research on how governments facilitate peace with rebel groups through power-sharing institutions (Hartzell and Hoddie 2003), electoral participation (Matanock 2016), or military integration (Glassmyer and Sambanis 2008). Given the post-war consequences of employing counterinsurgent militias during a war, it is important to understand the optimal conditions for governments to choose different militia management strategies (Staniland 2015). Failing to account for key structural, social, and political conditions in a post-war country can have devastating consequences for peace-building and state-building alike.

Explaining the Fates of Victorious Militias

We contend that the choice of different victorious militia management strategies undertaken by governments is influenced by the social relationship between these militias and the ruling elite as well as the organizational features of the standing military. These patterns are the result of a strategic dilemma between post-war ruling elites and the non-state groups that brought them to or kept them in power through civil war. Before explicating our argument in detail, however, we should first clarify that we focus on de facto militia integration and disintegration. De facto integration refers to the (new) government’s victorious militia partners actually being incorporated into the formal apparatus of the state’s official military. This is distinct from promises or agreements to integrate a group, which often fail but are almost always the focus of empirical research on rebel-military integration (Krebs and Licklider 2015). Moreover, integration processes sometimes take years to complete, and we focus on when the process has substantially altered the non-state organization and formally incorporated integrated units into the official military. For instance, the Policia Militar Ambulante paramilitary organization in Guatemala was not effectively integrated until two years after the civil war had ended. Disintegration for our purposes refers to the dismantling of militias via demobilization or forcible suppression or the severing of ties between the governments and group. A disintegrated militia group has a broken organizational infrastructure, rendering its ability to reorganize and rearm extremely difficult (Hartzell 2009). In other cases, governments formally terminate their relationship with the group and may even directly attack them (e.g. FDLR in the Democratic Republic of the Congo).

Ethnic Differences and Militia Disintegration

When a civil war has subsided, a dilemma arises between the (new) government and its armed militias over the distribution of power and/or resources (Zeigler 2016). The ruling elite seeks to implement its political and ideological state-building goals (Staniland 2015), and militias with incompatible interests may detract from those objectives, using their power to pursue independent goals that dilute the power of the ruling elite or even directly contradict them (Carey and Mitchell 2017). Without a common enemy to unite them and the redistribution of post-war power at stake, identity distinctions between the ruling elite and its non-state auxiliaries exacerbate mistrust in the post-war period through a process of asymmetric information, a shrinking of the shadow of the future, and commitment problems.

When ethnicity does not demarcate salient cleavages with respect to domestic power and resources, any agency problems in the post-war period are not reinforced by identity groups. Divergences in interests may still exist, as in all principal-agent relationships, but the security dilemma characterizing the post-war period is not delineated by ethnic differences that might otherwise enhance intergroup mistrust (Posen 1993). Militias in these settings may have also been created to be more permanent repressive forces (Ahram 2011). Moreover, these groups may have political or ideological alignments to the ruling elite, making their continued presence as auxiliaries after the conflict more feasible and perhaps beneficial to the state (Staniland 2015). Preferences to disintegrate these auxiliary forces is therefore far less likely than when the ruling elite and the militias have clear ethnic differences.

Similarly, when non-state forces are comprised of a similar ethnic composition as the ruling elite, there is a greater likelihood of aligned political interests, goals, and ideological perspectives between the ruling elite and armed militia than when the ruling elites and militia have distinct ethnic identities (Magid and Schon 2018; Abbs, Clayton, and Thomson 2020). Such interest alignment may reduce fears about shirking during the conflict, create incentives for loyal in-group behaviors, and increase the prospects of semiofficial ties (Carey and Mitchell 2017; Abbs, Clayton, and Thomson 2020). These groups also tend to be larger, more professional “paramilitary” groups and may serve additional roles for the regime such as counterbalancing against an otherwise independent official military (Carey and Mitchell 2016; Böhmelt and Clayton 2018). From the perspective of the principal (ruling elite), shared social ties with militias can ameliorate asymmetric information problems by increasing the credibility of assurances made by the militia and establishing clear expectations for in-group behavior (Fearon and Laitin 1996; Abbs, Clayton, and Thomson 2020). Lower uncertainty about intentions reduces the risks of maintaining even well-organized and militarily powerful co-ethnic militia allies after the civil war. From the perspective of the militia, this congruence of interests will incentivize the group to actively promote the state’s security goals. Doing so helps these groups generate benefits for themselves such as a regular income for members and enhanced domestic status. These incentives and expectations extend each actor’s respective time-horizon, thereby improving the prospects for long-term principal-agent cooperation and increasing the probability that the militia will endure as a post-war auxiliary force.

In contrast, militias that are ethnically distinct from a principal–which are the focus of this analysis–are often temporary organizations employed in “emergency” situations like civil conflict (Raleigh 2016; Abbs, Clayton, and Thomson 2020). These groups may be formally mobilized by elites from distinct ethnic groups, including groups that are otherwise excluded from power. They could even be comprised of civilian members of ethnic groups represented by the rebels (Lyall 2010; Kalyvas 2008; Abbs, Clayton, and Thomson 2020) or organized units that were previously part of a rebel organization (or government forces) and had defected (Otto 2018).¹ These militias in particular “are unlikely to be of sufficient size or loyalty to be…a viable long-term strategy” for stabilizing domestic security after the war (Abbs, Clayton, and Thomson 2020, 908). Absent a systematic and organized rebel threat, these militias typically are not well designed for post-war repressive purposes or coup-proofing (Böhmelt and Clayton 2018; Abbs, Clayton, and Thomson 2020). More importantly, the political interests and preferences of ethnically distinct militias often diverge from the ruling elite after the war has ended, leading to a “trust deficit” between principal and agent. The post-war ruling elite prefer to maximize their capacity to implement their state-building objectives and will be unlikely to share power with agents that could threaten these goals, particularly if they represent a rebel group’s constituency (Staniland 2015). Ethnically distinct militias thus fear being taken advantage of or suppressed by the post-war government despite having fought on the same side during the conflict. In the post-war period, therefore, ethnically distinct militias have incentives to find ways to enhance their own bargaining leverage, while post-war regimes have strong incentives to avoid incorporating them into regular politics (Staniland 2015).

These rational expectations of ethnically distinct ruling elites and unofficial militias engenders a process of mutual distrust among former allies. More specifically, incompatible interests (or the perception thereof) leads to an acute time-inconsistency problem between principal and agent in which the government’s ethnically distinct militias are “friendly” today but may emerge as a serious threat to the state’s security interests in the future. At the very least, when the principal and agent are comprised of distinct identity groups, problems of asymmetric information with respect to the political goals of each are exacerbated (Abbs, Clayton, and Thomson 2020). The principal in this case will lack information ex ante about the militia’s capabilities and intentions to comply with the state’s security directives (Carey and Mitchell 2017), and therefore will have few incentives to maintain any cooperative power-sharing arrangement after the immediate opposition threat has been terminated. If the state ignores these militias, they can become increasingly powerful and create competing power centers, thereby ensuring that governments “lose authority over large sections of territory” (Rotberg 2004, 6; Raleigh 2016). This undermines state authority and may even increase the likelihood of renewed violence (Steinert, Steinert, and Carey 2019). For these reasons, Reno (2011) observes that in post-Cold War Africa, “armed factions associated with past and present governments, conceived in part as instruments to bolster these governments, came to be the principal threat to their security” (p. 244).

These uncertainties lead to a severe commitment problem after a civil war has ended: ethnically distinct militias struggle to credibly promise to the government that they will not act against the state’s interests, leading the government to prefer eliminating the group rather than retain it during the post-war period. In fact, being aware of this dynamic, these militias will be more prone to renege on any agreement with the government in part because they anticipate that the state will defect from its side of the bargain as well. This increases the prospects for agency loss or defection as it may drive these militias to pocket resources they receive from patrons or from looting while at the same time striking their own deals or targeting rivals regardless of the strategic interests of the state. Governments, in turn, will find it increasingly difficult to restrain and control the actions of these militias if allowing them to operate freely.

The primary implication of this vicious cycle and shrinking of the shadow of the future is that after a war has ended and political power is to be redistributed, the ruling elites have strong incentives to dismantle their former militia partners if there are clear identity cleavages between them. Doing so, however, is neither straightforward nor without risk. Suppressing or demobilizing and disarming militias requires time, effort, and resources and may invite a violent retaliatory backlash if these groups fear that “disbanding and disarming…limits (their) capacity for self-defense” (Krebs and Licklider 2015, 103; Carey and Mitchell 2016).² However, disintegrating ethnically dissimilar militias is far less costly than dismantling militias with stronger social ties to political elites because the militia will have few advocates within the ruling coalition. In sum, these mutual commitment problems combined with the relatively low political costs of disintegrating ethnically distinct militias lead us to expect the following:

H1: Post-war governments will be more likely to disintegrate their non-state militia allies when the militia’s composition is ethnically distinct from that of the ruling elite.

Military Organizational Capacity and Militia Integration

If a government chooses not to disintegrate its militias, it will sometimes absorb them into the official military. This can be a useful strategy for co-opting certain groups (Staniland 2015), but integrating untrained, violent, and possibly unruly armed militants may also impair the military’s coherence, organizational structure and operational efficacy (Krebs and Licklider 2015; Martin 2018). Military integration is a costly process that requires the investment of resources as well as institutional buy-in from the regular military corps,³ and overcoming these hurdles still does not guarantee that integration will subdue these new units, let alone satisfy their independent interests (Krebs and Licklider 2015). For many states, after fighting a costly war, financial and political resources are scarce, and the costs of integration may be too great to immediately bear. This outcome should thus be less frequent than other militia management strategies in post-war countries, and the probability of effective integration should be primarily a function of the organizational capacity of the security apparatus as opposed to social cleavages between the ruling elites and militias. We argue that military integration is most likely at intermediate as opposed to low or high levels of military organizational capacity.

Militaries with low military organizational capacity are characterized by a shortage “of supplies and trained personnel,” which severely compromises their operational efficacy (Beckley 2010, 49; Cohen 1986). The security forces in these states also tend to be plagued by “poor tactical initiative, weak combined arms practices, [and] intelligence failures” that limit their combat abilities (Beckley 2010, 49; Tellis et al. 2000). Further, they are often organizationally underdeveloped and “face an acute shortage of the kind of managerial expertise needed” to operate efficiently (Cohen 1986, 45), which leads to poorly trained or disorderly units. Consequently, these forces are hampered by multiple structural, organizational, and strategic constraints that should reduce their ability to integrate militias.

First, in some cases, states with low military organizational capacity may not have adequate finances to effectively incorporate a substantial number of members from unofficial militias (Howe 2002; Tellis et al. 2000). While such states may provide some resources to integrated militia members, they may be unable to credibly promise to sustain funding in the long-run. In these cases, the state’s lack of credibility to deliver on any benefits to former militia members could mean that the expected gains from joining the military from the perspective of the militia may be outweighed by the chance to return to civilian life or secure power or financial resources through alternative, possibly illicit means. This reduces the ex ante incentives for the militias to join the armed forces or remain disciplined and loyal even if they were to be integrated (Koren and Mukherjee 2021). In the Central African Republic, for example, auxiliary militias anticipated little to “no reward of government or state largesse” from joining the official armed forces and thus “continued to loot from urban residents” to sustain themselves (Isaacs-Martin 2015, 1). Thus, auxiliary militias, regardless of their own capabilities, may prefer to remain independent rather than accepting more constraints within a weak military organization and the possibility of fewer material payoffs.⁴

Compounding these issues, the poor operational abilities that characterize weak military organizations constrain their ability to monitor and control untrained and possibly unruly militia members who join the official military (Gaub 2011; Martin 2018). Integrating non-state forces into these militaries can cause “deep dysfunction and fragmentation…low internal cohesion, imbalanced organizational structure[s], and [a] lack of accountability,” leading to a breakdown of an already weak security apparatus (Martin 2018, 528). Moreover, poor monitoring and subsequent indiscipline within the military can lead to agency slack among integrated militia units where, even after integration, former militia members refuse to comply with government or officer demands. Thus, weak military organizations have few incentives and little capability to integrate militias after a civil war has ended, and militias themselves have little to gain from being integrated even if they had the capability to pressure the government for integration (Koren and Mukherjee 2021).

In contrast to organizationally weak militaries, security forces characterized by high levels of organizational capacity “have sufficient strength in numbers, organization and capability to overcome its likely opponents” (Liddle 1999, 27; Cohen 1986). These larger militaries also tend to have significant organizational links between various units and are often led by entrenched military officers with a “bureaucratic and hierarchical” mindset who favor existing “rules, structure, predictability,…rank, authority, and exclusivity,” while opposing change to the military’s force structure (Holmberg and Alvinius 2019, 134; Howe 2002; Gaub 2011; Galvin 2018). These internal structural features are important for maintaining larger military organizations given their expansive capabilities and operations. Thus, from an organizational perspective, “the size of the military force is [an] important” feature of such militaries as “quantity has a quality all its own” (Tellis et al. 2000, 138).

Militaries with high levels of organizational capacity should be associated with a low probability of militia integration for at least three reasons. First, these states have little reason to further bolster the size of their official forces because their capabilities are already high. Second, the coercive capabilities of these strong militaries will induce governments to believe that they have enough of a military advantage to subdue their contemporary militia allies in the future if necessary. This does not mean that they will necessarily disintegrate these militias, however, because auxiliary forces can continue to serve other purposes after the war has ended (Carey, Colaresi, and Mitchell 2015; Carey and González 2020). Rather, militias in these states can be retained as unofficial auxiliaries because the ruling elites will have confidence that they can control or eliminate the militias in the future if need be.

Third, organizationally strong militaries often have an entrenched officer class that will staunchly resist any change to the military’s units and organizational structure (Holmberg and Alvinius 2019, 134; Howe 2002). These officers, who often have a formal background in military training and academies, may view non-state militias as amateurs that will impair the military’s preparedness and its capability to effectively respond to organized domestic threats. Attempts to integrate militias into these militaries will be viewed by these officers as an inflammatory action (even if the militia is co-ethnic) that can severely erode military discipline or exacerbate political tensions (Abbs, Clayton, and Thomson 2020). In countries like Pakistan and Nigeria, which have large, professional armies, the commanding officers have traditionally resisted the integration of auxiliary militias owing to concerns about their unprofessionalism and lack of discipline (Gaub 2011). Thus, although larger militaries may have more “organizational space” for integrating new units and can credibly deliver greater benefits to integrated militia members, they will also be resistant to dramatic organizational changes. Militias, for their part, could gain from integration into these militaries because of their ability to credibly deliver a steady flow of income, but these particularly strong militaries will often have far greater bargaining power than a state’s unofficial auxiliaries. The interests of these militaries, therefore, will often supersede those of the militia. Thus, the preferences of these large, organizationally strong militaries, combined with their greater coercive capability and bargaining leverage, reduce the odds of integrating militias.

We expect, however, that the probability of military integration should be greatest at intermediate levels of military capacity. These security organizations are moderately-sized, but also tend to be in a state of rapid evolution with respect to operational strength and organizational depth (Howe 2002; Beckley 2010). The force structure and number of units in these organizations are often in the midst of an upward trajectory (Tellis et al. 2000; Gaub 2011), making it more feasible for these forces to grow rapidly, develop greater “integrative capacity through structural and organizational specifics,” and develop new doctrines (Gaub 2011, 17; Howe 2002). Indeed, these armed forces have greater maneuvering room for organizational development, which provides them with the capability to incorporate new units from non-state armed groups. This is complemented by the fact that these militaries are less financially constrained than their weaker military counterparts. Access to sufficient resources ensures that the government can credibly provide material benefits for participation in the military. A steady income-stream will encourage non-state soldiers to join the official military and also incentivize them to not deviate from the state’s interests or goals, thereby decreasing internal agency slack.

Finally, these militaries have a reasonable, but not substantial, degree of coercive power. This will create enough uncertainty within the government about its ability to suppress unruly militias in the future to discourage it from allowing the militias to remain separate entities. In post-war environments when state-reconstruction is particularly important, managing these militias is crucial for creating political stability, but these intermediate militaries may not wish to incur the relatively high costs of suppressing them. From the perspective of the militia, the relative organizational openness and reasonably credible promises for material gains will lead the group to rationally expect that the state will honor its commitment to the integration process. Regardless of the capabilities of the militia itself, furthermore, both the government and militia will be likely to favor integration under these conditions.⁵ We therefore expect that

H2: Military capacity has a parabolic (inverse-U) relationship with the probability that a victorious militia is integrated into the military after a civil war.

Our primary hypotheses are derived from our central argument that the organizational capacity of the post-war military is the primary driving force behind the successful military integration of victorious militias, whereas social cleavages between ruling elites and militia allies are associated with a higher probability of militia disintegration. In the next section, we evaluate the independent effects of our two independent variables on the probability of each militia fate. However, some combinations of our independent variables may also amplify or moderate the effects of the other. Based on the arguments presented here, we expect that ethnic differences between the ruling elites and their victorious militias still severely and independently decreases the probability of military integration regardless of the level of military capacity. Our expectations for the interactive effects of ethnic differences and military organizational capacity are more complex. When military organizational capacity is low, ruling elites will fear that militia allies will take advantage or their empowered status in the post-war era if they are not disintegrated, but the weak capabilities of the military may prevent them from doing so. On the contrary, when ethnic differences are present and military organizational capacity is high, mistrust between the elites and militia will still be acute, but a strong military will be able to credibly monitor and deter militia malfeasance, thereby lowering the immediate need to demobilize the militia. Instead, the state can continue taking advantage of the political benefits of having unofficial auxiliary forces comprised of other ethnic groups with the foreknowledge that it can control the group rather than fear it. Finally, although states with militaries with moderate organizational capacity are generally expected to integrate militia allies after a war, ethnic differences can increase the probability that militias under these conditions will be disintegrated instead. These militaries may be more uncertain about their future abilities to monitor and control militias that may already have greater goal variances vis-à-vis the state, but they also have a reasonable level of capability (and incentive) to demobilize the militia immediately. Taken together, we expect that high military organizational capacity should make ethnically different militias less likely to be disintegrated, relative to other militaries, and that ethnic differences should increase the probability that militias are disintegrated when the military has moderate capacity, relative to when there are no ethnic differences. Given space constraints, we explicate these arguments in greater detail and test them empirically in the Supplemental Appendix.

Research Design

Non-state Victorious Militias and Sample Frame

Since we are interested in studying incentives for post-war governments to manage their militias in different ways, the relevant groups for our analysis are the non-state militias that fought on behalf of the party in power after the war. We define a victorious militia as armed, non-state organization that fought as part of the victorious side in the war and/or explicitly coordinated with or was subservient to the ruling elite in charge of the executive after the conflict has ended. To be included in our dataset, the militia group must have actually participated in the war effort, been explicitly part of or subservient to the leaders that took power or sustained their power through war, and been active in the final year of the war. Groups that are terminated prior to the end of the war are excluded because we are focusing on the unique political dynamics of the post-war environment. Since militias and other non-state forces can be used for a variety of purposes, we also kept our focus on those involved in insurgent or counterinsurgent operations by either directly fighting against the opposing side or actively repressing civilian constituencies that might otherwise support their opponents. More details about coding criteria are provided in the Supplemental Appendix.

Our sample frame consists of peace years after all civil conflicts ending between 1989 and 2014.⁶ We relied upon the UCDP conflict termination database to identify the relevant conflicts (and outcomes), which are defined by an inclusive threshold of twenty-five battle deaths per year (Kreutz 2010). However, using a dyadic unit of analysis (government-rebel-post-conflict year) would introduce a number of oddities in our sample frame. For instance, if two separate groups are fighting against the government at the same time and then only one of them negotiates a peace agreement, we cannot consider the subsequent years as post-conflict peace years because conflict is still ongoing with another group. This presents multiple difficulties for our purposes. First, victorious militias are defined as groups that fought for the post-war government (incumbents or rebels that took power), and this “victor” in one dyadic conflict should not necessarily be considered the “victor” of the larger civil war. Second, if a militia fought against multiple rebel groups, and these rebel groups are defeated in different years, a dyadic approach would require that the same militia enter our dataset for all years after the first rebel group was defeated and all years after all other groups were defeated. These duplicate militia-years would artificially deflate our standard errors and yield nonsensical results.

We resolved this problem by aggregating the dyadic conflicts in the UCDP data that were ongoing at the same time, thereby conceptualizing civil war more generally as “the rupture of state sovereignty” (Sambanis and Schulhofer-Wohl 2019) while still taking advantage of the inclusiveness of a low battle-death threshold. In doing so, we are able to more clearly identify true post-conflict peace years while preserving the militia-level information that is important for our analysis. After identifying the end of each civil war in this way, we create our militia-peace year panel dataset by including all years after the war has ended (including the final year of the war) until (a) the victorious militia terminates, (b) the year prior to the onset of another conflict in which the militia participates, or (c) 2014 if the militia does not engage in any other conflicts but was never dissolved or integrated.

To define our sample, we first identified which side’s ruling elites had power after the conflict had ended. If the regime during the conflict retained power after the conflict or did not lose power as a result of the rebels overthrowing the government, we included the pro-government militias that fought against the rebellion. In these cases, we began with the Pro-government Militia Database (PGMD; Carey, Mitchell, and Lowe 2013) and Relational Pro-Government Militia Database (RPGMD; Magid and Schon 2018). Using secondary sources online, news articles, NGO reports, scholarly articles, and academic books, we verified that each relevant group had some linkage with the government, determined whether the group was an active participant in the civil conflict, and gathered information about the nature and date of the group’s termination based on our coding criteria. We also expanded our search using secondary sources and academic works, which led us to include groups that were absent in these databases. If rebel leaders forcibly removed and replaced the executive as part of the conflict outcome or the rebels successfully seceded (through outright victory or as part of a peace agreement), we included the non-state forces that fought as part of the rebellion as well as any of their auxiliary forces. For instance, when the EPRDF in Ethiopia successfully ousted the Derg government in 1991, the non-state “victorious militias” in our sample include the EPRDF, the OLF (who coordinated with the TPLF-EPRDF in the late 1980s), and the People’s Militia forces that the EPRDF had converted as they advanced to Addis Ababa. We relied upon a similar milieu of resources to identify any additional subsidiary militia groups not included in the UCDP. Using these coding procedures, we identified ninety-six total conflicts in forty-nine different countries that had at least one victorious militia. Our final sample consists of 132 unique militias and 650 militia-peace years.

Dependent Variable

A non-state militia may be integrated into the military, disintegrate, or remain active in any given post-war year. To construct our dependent variable, we determined whether, when, and how each of our victorious militias terminated. The time-varying nature of our data is unique, as much of the literature on military integration and disarmament, demobilization, and reintegration (DDR) processes focus on the provisions in peace agreements rather than their actual implementation (Glassmyer and Sambanis 2008). We sought to measure de facto integration and dissolution rather than the promise to do so, which turns out to be important because many groups in our dataset terminate multiple years after the war has ended. In fact, of the victorious militias that terminated in one way or another, the average post-war survival time was approximately four years.

We coded a group as being integrated into the formal military apparatus if a substantial portion of the militia could and did join the official state military or other security forces without substantial limitations on promotion and if any quota agreed upon by warring parties was not severely limiting. Since integration is often a process rather than a discrete event, many cases of military integration take place over multiple years. We code the year of integration as the year in which the majority of the group was integrated, or otherwise the year in which the remainder of the group is no longer combat-ready or mobilized for war. We obtained this information from a variety of academic sources, NGO reports on peace agreement implementation, and news articles. Twenty-four militias in our dataset were integrated into the military sometime after the civil war ended.

Though many militias do integrate into the military, most are disintegrated. We define disintegration broadly: victorious militias are considered disintegrated in the year that the majority of the group is demobilized and therefore no longer battle-ready, or when a clear separation occurs between the (new) government and the militia group. In the latter instance, the government formally annuls any relationship with the group, and sometimes launches a military campaign to eliminate it. In most of the disintegration cases, however, the government oversees and implements a formal demobilization of the group. Fifty-five militias in our dataset were disintegrated. If the group was not disintegrated or militarily integrated in a given year, we code the group as ongoing until it terminates or a new conflict in which it fights begins.

Independent and Control Variables

To code our first independent variable of interest, we identified whether each militia has a clear religious or Ethnic Difference with the primary victor of the war. We considered a group to have an ethnic difference if the organization is comprised of or primarily recruits from a specific identity group than the victorious rebel group or post-war ruling government party or if we find that these two entities are religiously or culturally distinct. If a militia group is not ethnically or religiously homogenous, we code this variable with a 0.

Our second independent variable is the Military Organizational Capacity of a country in a given year. Since our focus is on the capabilities of militiaries and the associated bureaucratic and organizational features that tend to characterize military organizations of various sizes, we measure this variable with data on total military personnel (logged) from the Correlates of War National Material Capabilities (v.5.0) dataset (Singer 1988) and The Military Balance (2014). We include a linear and squared term to test the parabolic relationship we predict. Admittedly, militaries are complex, multidimensional organizations with multiple overlapping structural and cultural features that are nearly impossible to capture in a single measure (Beckley 2010; Cohen 1986; Howe 2002; Tellis et al. 2000). However, as we explain above, the size of a military organization also tends to correlate with other structural features that are relevant for testing our theory. Larger military organizations, for instance, require a massive professional and often rigidly structured bureaucratic administration to effectively manage and transmit information through the complex networks of units that make them up (Beckley 2010). While this is important for translating raw capabilities to efficient coercive action, “very large [military] organizations have a robust and powerful immune system” with respect to organizational change “due to distance and diffusion” as well as cultural, political, and practical tendencies among the officer class toward resisting change (Galvin 2018, 104; Holmberg and Alvinius 2019). We argue that these features can make integrating unofficial militias particularly challenging in larger militaries. Thus, for our purposes, “the size of the total force” is at least an appropriate “first cut” measure of our concept of interest.

We control for a variety of factors that could also affect the fates of militia groups after a war. First, we control for whether the government and rebels signed a peace agreement that included DDR or military Integration Provisions for rebel troops. These agreements often do not include specific provisions related to auxiliary militias, and indeed, militia groups outside the primary warring parties are rarely present during negotiations (Steinert, Steinert, and Carey 2019). However, these provisions may influence the post-war government to dissolve or integrate its non-state forces as well. Using information from the UCDP Peace Agreements dataset (Petterson and Öberg 2020), we created separate binary DDR and military Integration Provisions variables where a 1 signifies that the warring parties had agreed upon the respective provision in a peace agreement at the war’s end.⁷ In order to account for external third-party influences, we also control for foreign mediation efforts using a binary indicator based on data from the Civil War Mediation (CWM) dataset (De Rouen, Bercovitch, and Pospieszna 2011). We updated this variable to account for mediation efforts until 2014. We also control for the country’s GDP/capita (logged) using data from the World Bank (2019) as well as the country’s regime type. We consider a country a reasonably strong democracy if its Polity score is at least 6 and an autocracy if its Polity score is at most −6 (Marshall, Gurr, and Jaggers 2014).

Finally, we include a binary indicator signifying whether the rebels took power or successfully seceded. The ruling elite face principal-agent concerns with their militia allies regardless of whether they are incumbent elites or seized power through force. However, a new regime following a rebel takeover often needs to restructure the official security apparatus in a way that it can faithfully execute the security interests of the new government. This is particularly difficult when the standing military has been obliterated by war against the very elites now in power, as the security apparatus will be inefficient and have its own incentives to shirk responsibilities and prefer not to support the new post-war regime and its state-building goals. For this reason, we expect that, all else equal, former rebel leaders that occupy the post-war seats of power will be more likely to deal with the principal-agent problems described in our theory by integrating their victorious militias, particularly the main non-state rebel organization that defeated the previous government’s official military.

Method of Analysis

Our dependent variable is tripartite and categorical. We therefore employ multinomial probit (MNP) model, which allows us to estimate the effects of our independent variables on all three types of militia fates. We choose the MNP model because it does not assume the independence of irrelevant alternatives like the multinomial logit. We adjust for heteroskedasticity in the error term by clustering the standard errors by conflict and include a cubic transformation of the number of peace years to account for time dependence.

Results and Analysis

We report our main MNP specifications in Table 1, which provides coefficient estimates for the independent variables’ effects on the probability of disintegration or de facto military integration relative to the probability that the group continues to exist in each peace year. Model 1 (the first two columns of Table 1) is our baseline MNP specification. We add controls for regime type in Model 2, while Model 3 includes all the controls listed earlier.

Table 1.

Multinomial Probit Analysis of Victorious Militias by Termination Type.

	Model 1		Model 2		Model 3
Dependent Variable:	Disint.	Military Int.	Disint.	Military Int.	Disint.	Military Int.
Military Capacity (log)	0.125	4.5**	−0.084	4.819**	0.42	4.447**
	(0.46)	(1.663)	(0.502)	(1.587)	(0.506)	(1.411)
Military Capacity (log)²	−0.057	−0.703**	−0.026	−0.748**	−0.092	−0.684**
	(0.062)	(0.239)	(0.07)	(0.233)	(0.071)	(0.21)
Ethnic Difference	0.947**	−0.318	1.013**	−0.242	1.11**	−0.022
	(0.263)	(0.466)	(0.28)	(0.462)	(0.305)	(0.504)
Integration Provs.	−0.089	0.865	0.315	0.754	−0.057	0.729
	(0.263)	(0.606)	(0.504)	(0.647)	(0.561)	(0.688)
DDR	1.01**	−0.799	0.904*	−0.916	1.323**	−0.548
	(0.362)	(0.594)	(0.368)	(0.661)	(0.377)	(0.671)
Mediation	−0.567	−0.748	−0.63	−0.668	−0.668	−0.889
	(0.392)	(0.528)	(0.386)	(0.556)	(0.386)	(0.565)
Autocracy			−1.323*	0.472	−1.244	−0.05
			(0.652)	(0.492)	(0.787)	(0.547)
Democracy			0.371	−0.283	0.442	0.025
			(0.34)	(0.491)	(0.387)	(0.49)
Rebel Victor					1.109**	1.451*
					(0.403)	(0.568)
GDP/Cap. (log)					−0.088	−0.063
					(0.123)	(0.211)
Peace Year	−0.942*	−1.556	−1.017*	−1.465	−0.807	−1.844
	(0.441)	(2.403)	(0.467)	(2.383)	(0.557)	(2.51)
T2	0.15	0.644	0.156	0.652	0.104	0.778
	(0.085)	(0.896)	(0.086)	(0.876)	(0.099)	(0.937)
T3	−0.007	−0.08	−0.007	−0.083	−0.004	−0.095
	(0.004)	(0.1)	(0.004)	(0.097)	(0.004)	(0.105)
Intercept	−0.36	−7.041*	−0.022	−7.718**	−0.716	−6.956*
	(0.842)	(3.084)	(0.889)	(2.893)	(1.164)	(2.873)
N	543	543	543	543	504	504
Log-likelihood	−186.026**	−186.026**	−180.334**	−180.334**	−160.559**	−160.559**

Note: *p < 0.05; **p < 0.01. Cluster-adjusted robust standard errors reported in parentheses. Reference category for the dependent variable is continued existence of the militia.

Beginning first with our test of Hypothesis 1, all three models provide strong evidence that victorious militias are more likely to be disintegrated if there is a clear ethnic difference between the post-war ruling elite and the militia. In all three specifications, the coefficient estimate of Ethnic Difference represents a positive and statistically significant (at the 99 percent confidence level) predictor of militia disintegration. When we add controls for regime type (Model 2) and then Rebel Victory and GDP per capita, the magnitude of the coefficient increases. Figure 2A illustrates the predicted probabilities of militia disintegration when the militia does and does not have a significant ethnic difference from the victor, including the 95 percent confidence intervals of these probabilities.⁸ The figure clearly indicates that the predicted probability of disintegration is substantively negligible for victorious non-state militias that do not have a significant ethnic difference with the ruling elite, thereby confirming our theoretical intuition discussed earlier. By contrast, there is a substantial and statistically significant positive influence on the predicted probability of disintegration for victorious militias that do have a distinctive identity difference with the ruling elite: these militias are approximately three times more likely to be dismantled compared to groups that do not have a significant ethnic difference with the victor.

Figure 2.

Predicted Probabilities of Militia Disintegration and Integration. Figure 2A depicts the simulated predicted probability of militia dissolution and the associated 95 percent confidence intervals when an ethnic difference does and does not exist. Figure 2B shows the predicted probabilities and 95 percent confidence intervals of military integration associated with changes in military personnel. In both cases, all other variables are held at their medians.

The associations between various controls and the disintegration outcome provide some interesting empirical insights. Most importantly, ethnic differences significantly contribute to the probability of militia disintegration over and above the effect of any existing DDR provisions agreed upon by the warring parties. This suggests that victorious militias are not demobilized simply because of political pressures to resolve a conflict; rather, victorious parties elect to disintegrate groups that could pose a threat to the future of the regime or otherwise inflame existing social cleavages that could be used to mobilize another rebellion. DDR provisions in recent peace agreements, for their own part, are also positively associated with victorious militia disintegration. While not the focus of our study, this result does indicate that demobilization provisions in peace agreements are more than just window dressing.

Interestingly, autocracies appear to be less likely than “anocracies” to disintegrate their victorious militias, relative to allowing them to remain active at arm’s length from the regime. Rebel Victory is also a highly significant positive predictor of disintegration (relative to remaining active as an auxiliary) in Model 3. Upon further examination into the cases, many victorious militias that fought in support of victorious rebel leaders either preferred to demobilize rather than join the army, or the relationship between the new government and their non-state allies was severed due to disputes over the distribution of post-war power. Future work can build upon this result to better understand how rebels become rulers after winning a war. No other control variables are statistically significant.

The MNP results for the military integration outcome provide strong support for Hypothesis 2. In all three specifications, Military Capacity and Military Capacity-squared are positive and negative respectively, and each are each statistically significant at the 99 percent level. These coefficients reflect a strong inverted-U relationship between military capacity and the probability of successfully implemented military integration relative to continued activity as auxiliaries.

This bell-shaped relationship is illustrated in Figure 2B, which displays the simulated predicted probabilities (and 95 percent confidence intervals) of military integration at each level of military capacity from the MNP specification in Model 1. Consistent with our hypothesis, the probability of military integration is clearly negligible for states in the lowest and highest third of the distribution of our Military Capacity variable, but increases significantly and eventually peaks at intermediate levels of military capacity. These probabilities are statistically significant between about 2 and 4 on our logged Military Capacity scale, which represents approximately 24 percent of the data.

Model 3 also provides strong evidence that integration of non-state armed groups into the formal military apparatus is far more likely after rebels have seized power than when governments retain it. This effect is unsurprising, but important to discuss. Whereas government victors already have an institutional foothold after the war, rebels must often alter the security institutions after victory as a first step toward re-building the state’s security sector (Berhe 2017). This often involves reforming the formal military apparatus and including members of armed wing of the rebellion and its affiliate militias in a newly structured national army. Consistent with the principal-agent logic often used to model the relationship between states and militias, new ruling elites can expect less shirking and a lower risk of betrayal if the military has elements of the non-state groups that brought them to power, especially relative to the military they recently defeated (Martin 2018). In combination with the significant and positive relationship between Rebel Victory and militia disintegration, this result suggests that rebel non-state forces rarely function well as auxiliaries or repressive forces in the post-war environment: they must either be appeased and integrated or suppressed if political tension, social unrest, or renewed conflict is to be avoided (Steinert, Steinert, and Carey 2019).

Turning to the other controls, we find the presence of integration provisions for rebels has no statistically significant association with the military integration of victorious militias. We find this surprising given the result from our DDR provisions for the militia disintegration outcome, but integration provisions do not always include militia forces (especially “informal” groups), and integration tends to be more difficult to successfully implement even with an agreement. None of the other control variables, including Ethnic Differences, are significantly associated with military integration after including our independent variable(s) of interest. Clearly, the integration of post-war victorious militias is primarily driven by organizational constraints and incentives in the military rather than country-level or conflict-specific dynamics.

Robustness Checks

Our results are robust to various alternative model specifications and research designs. In Table 2 we report a few of these specifications, including models with controls for different conflict outcomes (Model 4), the number of militias and whether the militia has “semiofficial” status (Model 5), and other conflict-related controls like whether the war had more than 1,000 deaths, the conflict’s duration, and the strength of the rebels (Cunningham, Gleditsch, and Salehyan 2013) (Model 6). Our variables of interest remain statistically significant across all of these specifications. Our results for semiofficial groups are also interesting: given that informal groups are not formally recognized by the state, they are often used to carry out human rights atrocities, which may make them less palatable candidates for integration. Indeed, our analysis suggests that semiofficial militias tend to be retained longer or are integrated, whereas informal groups tend to dissolve after a war’s end. Alternatively, semiofficial groups tend to be larger and more well-organized relative to informal groups, so semiofficial status could be a rough proxy for militia capacity. Interpreted in this light, more capable militias are unsurprisingly more likely to survive after a war and may even integrate at higher rates, but it does not eliminate the persistent influence of military capacity and ethnic differences on militia termination types.

Table 2.

Multinomial Probit Robustness Checks—Victorious Militias by Termination Type.

	Model 4		Model 5		Model 6
Dependent Variable:	Disint.	Military Int.	Disint.	Military Int.	Disint.	Military Int.
Military Capacity (log)	0.031	4.574**	0.386	4.679*	1.207	8.57**
	(0.526)	(1.052)	(0.586)	(1.989)	(0.927)	(2.534)
Military Capacity (log)²	−0.04	−0.699**	−0.083	−0.721*	−0.255	−1.319**
	(0.073)	(0.223)	(0.086)	(0.293)	(0.145)	(0.381)
Ethnic Difference	0.944**	−0.268	0.908**	−0.303	0.844*	−0.088
	(0.302)	(0.463)	(0.323)	(0.459)	(0.383)	(0.628)
Integration Provs.	0.182	0.799	−0.132	1.169	−0.371	1.244
	(0.572)	(0.669)	(0.692)	(0.788)	(0.758)	(0.918)
DDR	1.238**	−0.632	1.164**	−0.438	1.039*	0.307
	(0.39)	(0.632)	(0.407)	(0.629)	(0.489)	(0.707)
Mediation	−0.81*	−1.085	−0.924*	−1.197*	−0.993*	−1.435
	(0.409)	(0.592)	(0.362)	(0.605)	(0.455)	(0.769)
Autocracy	−1.458	−0.346	−1.415	−0.75	−1.365	−1.468
	(0.798)	(0.624)	(0.865)	(0.659)	(0.951)	(0.932)
Democracy	0.481	−0.000	0.649	−0.732	−0.054	−1.134
	(0.419)	(0.576)	(0.453)	(0.585)	(0.551)	(0.615)
GDP/Cap. (log)	−0.108	−0.033	−0.217	0.034	−0.11	−0.533
	(0.127)	(0.222)	(0.13)	(0.231)	(0.247)	(0.418)
Outright Rebel Victor	1.637*	2.394*	1.674**	3.038*	1.52	4.254**
	(0.67)	(0.989)	(0.708)	(1.286)	(0.931)	(1.46)
Outright Gov. Victor	1.113	0.744	1.051	1.495	0.786	2.439*
	(0.633)	(0.878)	(0.689)	(1.176)	(0.887)	(1.01)
Agreement/Ceasefire	0.462	0.798	0.261	1.87	−0.257	2.926**
	(0.509)	(0.718)	(0.577)	(1.085)	(0.673)	(0.998)
Number Militias			0.015	−0.347*	0.273	−0.452*
			(0.095)	(0.159)	(0.157)	(0.2)
Semiofficial			−0.768*	0.655	−1.04*	1.254*
			(0.318)	(0.552)	(0.41)	(0.523)
War Intensity					1.711*	−0.523
					(0.669)	(0.959)
War Duration					−0.004	−0.054
					(0.039)	(0.051)
Separatism					0.311	1.397
					(0.464)	(0.715)
Rebel Strength (log)					−0.308	−0.312
					(0.223)	(0.238)
Intercept	−0.157	−8.234*	0.389	−8.406	0.937	−10.174**
	(1.22)	(3.749)	(1.249)	(4.659)	(2.865)	(3.807)
N	504	504	500	500	454	454
Log-likelihood	−158.047**	−158.047**	−148.102**	−148.102**	−126.754**	−126.754**

Note: *p < 0.05; **p < 0.01. Cluster-adjusted robust standard errors reported in parentheses. Reference category for the dependent variable is continued existence of the militia. Cubic time polynomials omitted to preserve space.

In the Supplemental Appendix, we also lagged our military personnel variable to alleviate concerns about endogeneity; re-coded our ethnic differences variable to also include ideological differences; and dropped certain years and observations that could be disproportionately influential. Econometrically, we replicated our main specifications using multinomial logistic regressions as well as separate probits and rare events logits. We also reconstructed our dataset to comprise rebel-militia dyads and controlled for other ongoing conflicts. None of these alterations led to substantial changes in our main results.

Conclusion

When are victorious non-state militias more likely to be integrated into the state’s official security forces or, alternatively, dismantled after a civil war has ended? We posit that ethnic incongruence between the (new) ruling elite and the militia leads to a higher probability of militia disintegration, whereas the probability of military integration is largely influenced by the incentives and constraints imposed on a government by the organizational capacity of the official military. Results from multiple statistical models provide robust support for our predictions. Our findings provide a variety of important contributions.

First, there is little extant systematic research on how governments tame or manage their victorious supportive militias outside of war contexts (Staniland 2015). Our analysis takes an important step forward by statistically evaluating when governments opt to disintegrate or integrate these groups. This is important because militias in many contexts are used for repressive purposes even after a civil war has ended, and a failure to effectively manage them (or dissolve them) can lead to continual human rights violations (Carey and González 2020) and significant threats to future domestic stability (Steinert, Steinert, and Carey 2019; Reno 2011). We argue that social and organizational constraints motivate states to refrain from terminating these groups despite the competitive politics that can emerge after a civil war even among allies (Zeigler 2016). Second, we find strong evidence for our theoretical claim that governments use the ethnic identity of their militia partners as a heuristic to assess their trustworthiness and operational value after a war has ended. Our findings suggest that the relationship between governing elites and ethnically distinct militias will be plagued with acute commitment problems in the post-war period, which has important implications for how and why states reorganize their security forces and repressive apparatus when conflict violence has subsided (Ahram 2011). Empirically, we also contribute new data on victorious militias active in civil wars and their post-war fates. Building on existing datasets like the PGMD and RPGMD, we identified new militias on either side of multiple conflicts. More importantly, we identified whether, how, and when each of these militias terminated, allowing us to further depart from most existing empirical work on armed group termination that focuses on rebel group termination and agreements to integrate or disintegrate them rather than when termination actually occurs.

Our results offer a guide to peacemakers for understanding the optimal options for managing certain militias after a war has ended. Attempts to integrate a militia, for instance, may be unsuccessful if the organizational features of the official military are not suited for such a difficult endeavor. Understanding and aligning the interests of both militias and the ruling elites, particularly if there are significant identity cleavages between them, is essential for successful demobilization or integration of these auxiliary forces. We also question extant policy and empirical analyses that either ignore victorious militias or consider them to be the same as rebel groups. The optimal strategy for dealing with opposition groups after a war has ended is not necessarily the same for dealing with allied militias. Failing to acknowledge these differences can lead to ineffective peace-building strategies.

Nevertheless, more work needs to be done to fully comprehend the post-war fates of militias. First, our data capture events of successful integration, but formal theoretical work could also provide interesting insights into when states will attempt to integrate their militias but fail. Second, conceptual and empirical research should delve into the relationships between militias and civilian constituencies. For instance, future work could investigate how militias formed from the ethnic constituencies of rebel groups may play different roles from other militias that share few social ties to the ruling elites (Abbs, Clayton, and Thomson 2020; Lyall 2010). Finally, future work must expand on the roles of militias in the post-war period and whether disintegration or integration can effectively promote enduring post-war peace.

Supplemental Material

Supplemental Material, sj-zip-1-jcr-10.1177_0022002721995528 - Security Consolidation in the Aftermath of Civil War: Explaining the Fates of Victorious Militias

Supplemental Material, sj-zip-1-jcr-10.1177_0022002721995528 for Security Consolidation in the Aftermath of Civil War: Explaining the Fates of Victorious Militias by Brandon Bolte, Minnie M. Joo and Bumba Mukherjee in Journal of Conflict Resolution

Supplemental Material

Supplemental Material, sj-zip-2-jcr-10.1177_0022002721995528 - Security Consolidation in the Aftermath of Civil War: Explaining the Fates of Victorious Militias

Supplemental Material, sj-zip-2-jcr-10.1177_0022002721995528 for Security Consolidation in the Aftermath of Civil War: Explaining the Fates of Victorious Militias by Brandon Bolte, Minnie M. Joo and Bumba Mukherjee in Journal of Conflict Resolution

Footnotes

Acknowledgments

We would like to thank Ore Koren for helpful guidance and insights at the early stage of this project. We would also like to thank Nguyen Huynh, Neema Esfandi, and Annabelle Bardenheier for their excellent research assistance.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship and/or publication of this article: The authors are grateful for generous funding and support from the College of the Liberal Arts at Penn State University.

ORCID iDs

Brandon Bolte

Minnie M. Joo

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Abbs

Luke

Clayton

Govinda

Thomson

Andrew

. 2020. “The Ties that Bind.” Journal of Conflict Resolution 64 (5): 903–32.

Ahram

Ariel

. 2011. Proxy Warriors. Palo Alto, CA: Stanford University Press.

Beckley

Michael

. 2010. “Economic Development and Military Effectiveness.” Journal of Strategic Studies 33 (1): 43–79.

Berhe

Mulugeta Gebrehiwot

. 2017. “Transition from War to Peace.” African Security Review 26 (2): 143–60.

Böhmelt

Tobias

Clayton

Govinda

. 2018. “Auxiliary Force Structure.” Comparative Political Studies 51 (2): 197–237.

Carey

Sabine

Colaresi

Michael

Mitchell

Neil

. 2015. “Governments, Informal Links to Militias, and Accountability.” Journal of Conflict Resolution 59 (5): 850–76.

Carey

Sabine

Colaresi

Michael

Mitchell

Neil

. 2016. “Risk Mitigation, Regime Security, and Militias.” International Studies Quarterly 60 (1): 59–72.

Carey

Sabine

González

Belén

. 2020. “The Legacy of War.” Conflict Management and Peace Science. doi:10.1177/0738894219899006.

Carey

Sabine

Mitchell

Neil

. 2016. “Pro-government Militias and Conflict.” In Oxford Research Encyclopedia of Politics. doi:10.1093/acrefore/9780190228637.013.33.

10.

Carey

Sabine

Mitchell

Neil

. 2017. “Progovernment Militias.” Annual Review of Political Science 20:127–47.

11.

Carey

Sabine

Mitchell

Neil

Lowe

Will

. 2013. “States, the Security Sector, and the Monopoly of Violence.” Journal of Peace Research 50 (2): 249–58.

12.

Clayton

Govinda

Thomson

Andrew

. 2016. “Civilianizing Civil Conflict.” International Studies Quarterly 60 (3): 499–510.

13.

Cohen

Eliot

. 1986. “Distant Battles.” International Security 10 (4): 143–71.

14.

Cunningham

David

Gleditsch

Kristian Skrede

Salehyan

Idean

. 2013. “Non-state Actors in Civil Wars.” Conflict Management and Peace Science 30 (5): 516–31.

15.

Rouen

Jr Jacob Bercovitch

Karl

Pospieszna

Paulina

. 2011. “Introducing the Civil Wars Mediation (CWM) Dataset.” Journal of Peace Research 48 (5): 663–72.

16.

Fearon

James

Laitin

David

. 1996. “Explaining Interethnic Cooperation.” American Political Science Review 90 (4): 715–35.

17.

Galvin

Tom

. 2018. Leading Change in Military Organization. Carlisle, PA: United States Army War College.

18.

Gaub

Florence

. 2011. Military Integration after Civil Wars. New York: Routledge.

19.

Glassmyer

Katherine

Sambanis

Nicholas

. 2008. “Rebel-military Integration and Civil War Termination.” Journal of Peace Research 45 (3): 365–84.

20.

Hartzell

Caroline

. 2009. “Settling Civil Wars.” Conflict Management and Peace Science 26 (4): 347–65.

21.

Hartzell

Caroline

Hoddie

Matthew

. 2003. “Institutionalizing Peace.” American Journal of Political Science 47 (2): 318–32.

22.

Holmberg

Arita

Alvinius

Aida

. 2019. “How Pressure for Chang Challenge Military Organizational Characteristics.” Defence Studies 19 (2): 130–48.

23.

Howe

Herbert

. 2002. Ambiguous Order. Boulder, CO: Lynne Rienner.

24.

Isaacs-Martin

Wendy

. 2015. “The Motivations of Warlords and the Role of Militias in the Central African Republic.” African Center for the Constructive Resolution of Disputes. Accessed April 15, 2019. https://www.accord.org.za/conflict-trends/the-motivations-of-warlords-and-the-role-of-militias-in-the-central-african-republic/.

25.

Jentzsch

Corinna

Kalyvas

Stathis

Schubiger

Livia

. 2015. “Militias in Civil Wars.” Journal of Conflict Resolution 59 (5): 755–69.

26.

Kalyvas

Stathis

. 2006. The Logic of Violence in Civil War. New York: Cambridge University Press.

27.

Kalyvas

Stathis

. 2008. “Ethnic Defection in Civil War.” Comparative Political Studies 41 (8): 1043–68.

28.

Koren

Ore

Mukherjee

Bumba

. 2021. “Militias After the War.” Unpublished manuscript. Bloomington, IN: Indiana University.

29.

Krebs

Ronald

Licklider

Roy

. 2015. “United They Fall.” International Security 40 (3): 93–138.

30.

Kreutz

Joakim

. 2010. “How and When Armed Conflicts End.” Journal of Peace Research 47 (2): 243–50.

31.

Liddle

R. William

. 1999. “Indonesia’s Unexpected Failure of Leadership.” In The Politics of Post-Suharto Indonesia, edited by Schwarz

Adam

Paris

Jonathan

, 16–39. New York: Council on Foreign Relations.

32.

Lyall

Jason

. 2010. “Are Coethnics More Effective Counterinsurgents?” American Political Science Review 104 (1): 1–20.

33.

Magid

Yehuda

Schon

Justin

. 2018. “Introducing the African Relational Pro-government Militia Dataset (RPGMD).” International Interactions 44 (4): 801–32.

34.

Marshall

Monty

Gurr

Tedd

Jaggers

Keith

. 2014. “Polity IV Project.” Accessed April 5, 2019. http://www.systemicpeace.org/inscrdata.html.

35.

Martin

Philip

. 2018. “Security Sector Reform and Civil-military Relations in Postwar Cote d’Ivoire.” African Affairs 117 (468): 522–33.

36.

Matanock

Aila

. 2016. “Using Violence, Seeking Votes.” Journal of Peace Research 53 (6): 845–53.

37.

Mitchell

Neil

Carey

Sabine

Butler

Christopher

. 2014. “The Impact of Pro-government Militias on Human Rights Violations.” International Interactions 40 (5): 812–36.

38.

Otto

Sabine

. 2018. “The Grass is Always Greener?” Journal of Conflict Resolution 62 (7): 1459–88.

39.

Peic

Goran

. 2014. “Civilian Defense Forces, State Capacity, and Government Victory in Counterinsurgency Wars.” Studies in Conflict & Terrorism 37 (2): 162–84.

40.

Petterson

Therese

Öberg

Magnus

. 2020. “Organized Violence, 1989-2019.” Journal of Peace Research 57 (4): 597–613.

41.

Posen

Barry

. 1993. “The Security Dilemma and Ethnic Conflict.” Survival 35 (1): 27–47.

42.

Raleigh

Clionadh

. 2016. “Pragmatic and Promiscuous.” Journal of Conflict Resolution 60 (2): 283–310.

43.

Reno

William

. 2011. Warfare in Independent Africa. New York: Cambridge University Press.

44.

Rotberg

Robert

. 2004. “The Failure and Collapse of Nation-states.” In When States Fail, edited by Rotberg

Robert

, 1–50. Princeton, NJ: Princeton University Press.

45.

Sambanis

Nicholas

Schulhofer-Wohl

Jonah

. 2019. “Sovereignty Rupture as a Central Concept in Quantitative Measures of Civil War.” Journal of Conflict Resolution 63 (6): 1542–78.

46.

Singer

David

. 1988. “Reconstructing the Correlates of War Dataset on Material Capabilities of States, 1816-1985.” International Interactions 14 (2): 115–32.

47.

Staniland

Paul

. 2015. “Militias, Ideology, and the State.” Journal of Conflict Resolution 59 (5): 770–93.

48.

Steinert

Christoph

Steinert

Janina

Carey

Sabine

. 2019. “Spoilers of Peace.” Journal of Peace Research 56 (2): 249–63.

49.

Tellis

Ashley

Bially

Janice

Layne

Christopher

McPherson

Melissa

. 2000. Measuring National Power in the Postindustrial Age. Santa Monica, CA: RAND Corporation.

50.

The Military Balance. 2014. International Institute for Strategic Studies. New York: Routledge.

51.

World Bank. 2019. “GDP Per Capita (Current USD).” Accessed April 5, 2019. https://data.worldbank.org/indicator/ny.gdp.pcap.cd.

52.

Zeigler

Sean

. 2016. “Competitive Alliances and Civil War Recurrence.” International Studies Quarterly 60 (1): 24–37.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.26 MB

0.06 MB

0.00 MB